Senior Lecturer at University of Lille, ENSAIT, GEMTEX, France.
Senior Scientific Advisor at UDA & IAD
Senior Assoc. Prof. Kim Phuc Tran
Kim Phuc Tran is currently a Senior Associate Professor (Maître de Conférences HDR, equivalent to a UK Reader) of Artificial Intelligence and Data Science at the ENSAIT and the GEMTEX laboratory, University of Lille, France. He received an Engineer's degree and a Master of Engineering degree in Automated Manufacturing. He obtained a Ph.D. in Automation and Applied Informatics at the University of Nantes, and an HDR (Doctor of Science or Dr. habil.) in Computer Science and Automation at the University of Lille, France. His research deals with Real-time Anomaly Detection with Machine Learning with applications, Decision Support Systems with Artificial Intelligence, and Enabling Smart Manufacturing with IIoT, Federated learning, and Edge computing. He has published more than 64 papers in peer-reviewed international journals and proceedings of international conferences. He edited 3 books with Springer Nature and Taylor & Francis. He is the Associate Editor, Editorial Board Member, and Guest Editor for several international journals such as IEEE Transactions on Intelligent Transportation Systems and Engineering Applications of Artificial Intelligence. He has supervised 9 Ph.D. students and 3 Postdocs. In addition, as the project coordinator (PI), he is conducting 1 regional research project about Healthcare Systems with Federated Learning. He has been or is involved (co-PI or member) in 8 national and European projects. He is an expert and evaluator for the Public Service of Wallonia (SPW-EER), Belgium and the Natural Sciences and Engineering Research Council of Canada. He received the Award for Scientific Excellence (Prime d’Encadrement Doctoral et de Recherche) given by the Ministry of Higher Education, Research and Innovation, France for 4 years from 2021 to 2025 in recognition of his outstanding scientific achievements.
From 2017 until now, he has been the Senior Scientific Advisor at Dong A University and the International Research Institute for Artificial Intelligence and Data Science (IAD), Danang, Vietnam where he has held the International Chair in Data Science and Explainable Artificial Intelligence.
Title: Designing ECG monitoring healthcare system with Federated Learning and Explainable AI
Deep learning plays a vital role in classifying different arrhythmias using electrocardiography (ECG) data. Nevertheless, training deep learning models normally requires a large amount of data and can lead to privacy concerns. Unfortunately, a large amount of healthcare data cannot be easily collected from a single silo. Additionally, deep learning models are like black-box, with no explainability of the predicted results, which is often required in clinical healthcare. This limits the application of deep learning in real-world health systems. To address the above-mentioned challenges, we proposed a privacy-preserving, efficient and interpretable/explainable AI-based end-to-end framework to address the limitations of deep learning applications for EEG signal classification. Firstly, we proposed a CNN-based autoencoder in a federated architecture to denoise the raw ECG signal from patients. When trained on the baseline dataset, The proposed autoencoder provided an excellent reconstruction of the raw input signals and improved the overall performance when applied in federated settings. Secondly, we proposed a new classifier for ECG signals. When the classifier was trained in federated settings it was able to improve the overall classification performance of the edge devices. Moreover, the experimental results on the baseline database revealed that the proposed framework achieved outperformed existing algorithms, including both centralized and federated ones. Furthermore, we extended the usability of our framework by providing a novel explainable module on top of the classifier, whose usefulness is visually demonstrated by showing that clinically meaningful heartbeat segments of ECG signals are indeed behind the classification results. Additionally, we also proposed a communication cost reduction method, which can significantly reduce communication costs while increasing the level of privacy protection of users’ ECG data against the global server. Hence, the proposed framework shows its applicability by providing many desirable properties including interpretability, privacy protection, communication cost reduction, and high accuracy in classification. Such a combination of such properties does not hold for other existing solutions, therefore making the proposed framework a unique solution for real-world healthcare applications where ECG signal classification is an important task. Eventually, the proposed framework will encourage (1) more healthcare data owners to participate in training a good machine learning model for patients and health professionals, with fewer privacy concerns, (2) more accurate diagnostic assistance in places with scarce access to cardiologists or healthcare facilities, (3) more interpretable classification results that can be used to identify new potential patterns leading to trigger heart arrhythmias. Hence, the proposed framework has great potential to be added to hospital computer software platforms to support the work of health professionals and ultimately reduce mortality and save human lives.
Presentation document, download at: